Spatial metaphors in thinking about other people

Transcription

1 VISUAL COGNITION, Spatial metaphors in thinking about other people Davood G. Gozli a, Penelope Lockwood b, Alison L. Chasteen b and Jay Pratt b a Department of Psychology, University of Macau, Macau S.A.R., People s Republic of China; b Department of Psychology, University of Toronto, Toronto, Canada ABSTRACT Spatial metaphors contribute to our capacity for abstract thought. Consistent with this idea, it has been shown that processing semantic information (related to valence, power, etc.) can bias performance in a spatial task. Advancing this line of work, the present study examined whether spatial metaphors have a role in thinking about other people. Participants read short vignettes about academic performance, health or social life, which described students in superior and inferior states. In Experiment 1, after reading each vignette, participants were explicitly asked to assign a location to each protagonist using a pen-and-paper task. Findings from this experiment provided initial indication that thinking about the protagonists could recruit spatial metaphors. In Experiments 2 and 3, each vignette was immediately followed by an implicit test of spatial association. In Experiment 2, participants performed a name-recognition task in response to the protagonists names presented above or below the central fixation. In this experiment, metaphorical congruency facilitated performance. In Experiment 3, participants were presented with names at central fixation, followed by a visual discrimination target ( X / O ) above or below fixation. In this experiment, metaphorical congruency interfered with performance. The diverging patterns of results are explained in terms of the conjunction and separation of the conceptual and perceptual components of the recognition task, respectively, in Experiments 2 and 3. Overall, the findings support the role of spatial metaphors in thinking about other people and, more generally, for the spontaneous use of space in conceptual processes. ARTICLE HISTORY Received 18 April 2017 Accepted 19 February 2018 KEYWORDS Conceptual metaphor; embodied semantics; social cognition Our sense of space is integral to our capacity for perception and action (e.g., Baldauf & Deubel, 2008; Desimone & Duncan, 1995; Deubel & Schneider, 1996), and it has been argued that the sense of space continues to play a role in higher cognition, particularly in our understanding of abstract concepts (e.g., Boroditsky, 2000; Casasanto, 2017; Lakoff & Johnson, 1999; Meier & Robinson, 2004). Empirical investigation of this idea has involved examining the effect of concepts on various measures of sensorimotor bias, indicating that concepts automatically activate their associated spatial codes (e.g., Chasteen, Burdzy, & Pratt, 2010; Fischer, Castel, Dodd, & Pratt, 2003; Gozli, Chasteen, & Pratt, 2013a; Gozli, Chow, Chasteen, & Pratt, 2013b; Gozli, Pratt, Martin, & Chasteen, 2016; Marmolejo- Ramos, Montoro, Elosúa, Contreras, & Jiménez- Jiménez, 2014; Meier & Robinson, 2004; Santiago, Lupáñez, Pérez, & Funes, 2007; Sasaki, Yamada, & Miura, 2016; Schubert, 2005; Wang, Lu, & Lu, 2016; Xie, Wang, & Chang, 2014, 2015; Zhang, Hu, Zhang, & Wang, 2015). In the present study, we asked whether spatial metaphors can be engaged spontaneously as we think about other people, particularly people we read about for the first time. Before providing the details of the present study, it is helpful to begin with an overview of two types of experimental tasks that have been employed in studying the role of spatial metaphors in conceptual understanding. Among the commonly used methods for examining space concept associations are variants of the spatial Stroop task (Lu & Proctor, 1995) and the attentional cueing task (Posner, 1980). In a spatial Stroop task, participants are presented with words whose physical location could be compatible or incompatible with their meaning. For instance, words LEFT and YESTERDAY are expected to receive faster responses when appearing at the left of fixation, compared to when they appear at the right (e.g., Huffman & Pratt, 2016; Weger & Pratt, 2008; Zwaan & Yaxley, 2003). Attentional cueing tasks, on the other hand, involve presenting participants with a sequence of two events: a cue and a target. What is manipulated is CONTACT Davood G. Gozli gozli@umac.mo 2018 Informa UK Limited, trading as Taylor & Francis Group

2 2 D. G. GOZLI ET AL. the compatibility between the meaning of the cue (e.g., YESTERDAY ) and the physical location of the target (e.g., Chasteen et al., 2010; Fischer et al., 2003; Gozli et al., 2013a). The two effects have both been explained by presupposing that concepts can activate their spatial associations which, depending on the concept s match or mismatch with a to-be-attended physical location, impact processing efficiency. Meier and Robinson (2004) used both types of tasks in examining the role of spatial metaphors in understanding valence. In their spatial Stroop task, they presented single words above/below fixation and instructed participants to categorize the words as positive or negative. They found positive categorization to be faster above fixation, while negative categorization was faster below fixation (see Lakens, 2012). We should note that, in the spatial Stroop task, spatial orienting (above/below) and categorization (positive/ negative) are aspects of a single task. A participant could orient above fixation to categorize the word HAPPY, in which case the positive valence and up are two features of a single event, which is why the compatibility between the two features facilitates performance (Hommel, Müsseler, Aschersleben, & Prinz, 2001). In their attentional cueing task, Meier and Robinson (2004) presented a word at fixation, which again participants were required to categorize as positive or negative. After the word, a target letter ( p or q ) was presented above or below fixation, and participants had to identify this letter target. The authors found that after a positive word, letter identification was faster above fixation; after a negative word, identification was faster below fixation. These results indicate that valence can bias attention toward the metaphorically compatible location. These interactions between spatial and conceptual processing have been observed across many studies (e.g., Ansorge, Khalid, & Koenig, 2013; Goodhew, McGaw, & Kidd, 2014; Gozli et al., 2013a, 2013b, 2016; Lu, Zhang, He, Zheng, & Hodges, 2014; Quadflieg et al., 2011; Schubert, 2005; Taylor, Lam, Chasteen, & Pratt, 2015; Zanolie et al., 2012). Some studies have demonstrated that the cue target compatibility effect depends on how the cue is processed, which suggests some degree of flexibility over whether a concept can activate a spatial metaphor (e.g., Dodd, Van der Stigchel, Leghari, Fung, & Kingstone, 2008; Torralbo, Santiago, & Lupiáñez, 2006; Zanolie & Pecher, 2014). It is important to note that, in an attentional cueing task, the cue (i.e., the concept with a spatial association) and the target (e.g., target letter p / q ) are two separate events and, therefore, responding to p above fixation and categorizing the word HAPPY involve two events that share a single feature (up) in common (Hommel, 2004; see also, Amer, Gozli, & Pratt, 2017; Boulenger et al., 2006; Gozli & Pratt, 2011; Sato, Mengarelli, Riggio, Gallese, & Buccino, 2008). Assuming that attending to each perceptual event involves integration of its constituent features, having a feature in common means that the two events cannot be processed at the same time (e.g., Hommel et al., 2001; Treisman & Gelade, 1980). That is why cue target compatibility can interfere with performance, particularly when the events have non-overlapping features, appear in close temporal proximity and each require focused attention (Amer et al., 2017; Estes, Verges, & Adelman, 2015; Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a; Lachmair, Ruiz Fernández, & Gerjets, 2016; Ostarek, Ishag, Joosen, & Huettig, in press; Ostarek & Vigliocco, 2017). It should be noted that the landmark study by Meier and Robinson (2004) found facilitation, not interference. Here there was a relatively long delay between the presentation of the cue and the presentation of the target, with the two events requiring separate responses (participants first evaluated the cue at fixation as positive/negative, and only then were presented with a target above/below fixation). Separate responses to the cue and the target provide a sufficiently long cue target delay, such that the cue s spatial feature is no longer occupied by the cuerelated response when encountering the target. At the same time, the cue s spatial feature remains sufficiently active to bias performance in favour of the compatible target location. This is thought to be a key reason why Meier and Robinson observed facilitation with cue target compatibility (Gozli et al., 2013a; Lachmair et al., 2016). Previous research has examined the effect of preexisting metaphorical associations (e.g., Meier & Robinson, 2004), or the acquisition of new associations via repeated exposure (Dolscheid, Shayan, Majid, & Casasanto, 2013; Girardi & Nico, 2017; Gozli, Moskowitz, & Pratt, 2014), but not the application of an existing metaphor to new situations. If spatial metaphors are involved in abstract thought, then they should

3 VISUAL COGNITION 3 also be recruited on the fly, as we think about new situations and people. Assuming that aspects of a new situation are interpreted in a metaphorical manner similar to a single concept such as valence, the corresponding spatial metaphors should be activated. This question is the focus of the present study. On the basis of literature on spatial associations with concepts (e.g., God/Devil, good/bad, happy/sad being associated with up/down; Chasteen et al., 2010; Gozli et al., 2013a; Meier & Robinson, 2004), we reason that thinking about a person in a superior state should be more readily associated with up/ above, while someone in an inferior state should be more readily associated with down/below. In the present article, we use the terms superior and inferior to refer to a person s academic success/ failure, social relatedness/isolation and good/bad health. We predicted that subsequently remembering people in those states would activate spatial associations that would, in turn, influence performance in a spatial orienting task. Before testing our hypothesis using the spatial Stroop (Experiment 2) and attentional cueing (Experiment 3) paradigms, we needed to generate vignettes about people in superior and inferior states, in order to gather some evidence in favour of the idea thinking about others can activate spatial representations. To do this, in our first experiment we asked participants to read a set of vignettes, and we asked them to place each protagonist on a two-dimensional space by marking the initial of each name (e.g., P for Paul ) on a sheet of paper (horizontal and vertical midlines were drawn on the paper). Thus, participants were asked to explicitly assign names of the protagonists to locations within a spatial frame of reference. To preview the findings, we found explicit assignment of names to locations was consistent with the metaphorical association: on average, names of superior protagonists were assigned to locations to the right and above the centre, while the names of inferior protagonists were assigned to locations to the left and below the centre. Having found that our vignettes produced the expected spatial biases in an explicit localization task, we used the same vignettes in the following two experiments. In Experiment 2, participants completed a name recognition task. Following the logic of the spatial Stroop paradigm, on each trial, a single name appeared above or below fixation (i.e., metaphorically compatible or incompatible with their superior/inferior state), and participants were instructed to report whether the name was presented in the preceding vignette. In Experiments 2 4, we expected that recognition of protagonists would activate the associated spatial metaphors, e.g., superior-is-up. In Experiment 2, spatial orienting and name-recognition were part of one task. On each trial, a name was presented above or below fixation (compatible or incompatible with its metaphorical association) and participants responded whether they recognized the name from the preceding vignette. Thus, for Experiment 2, we expected faster responses when names were presented at metaphorically compatible locations (e.g., name of a superior protagonist above fixation), relative to when names appeared at incompatible locations (e.g., name of a superior protagonist below fixation). In Experiments 3 4, each trial consisted of performing two sub-tasks. The two-subtasks consisted of (a) name-recognition at fixation and (b) responding to a peripheral visual target (e.g., X / O ). Participants were instructed to respond the visual target only when the name belonged to the preceding vignette and withhold responding otherwise. In referring to the present experiments, we have reserved the term recognition to refer to the treatment of protagonist names, and we have used the terms discrimination (Experiment 3) and detection (Experiment 4) when referring to responses to the single-lettered, peripheral visual target ( X / O ) above/below fixation. Recognition of a centrally presented name, on the one hand, and responding to a peripheral visual target ( X or O ), on the other hand, were part of two sub-tasks in Experiments 3 4. We reasoned that the first sub-task, i.e., name recognition, could interfere with the second sub-task, i.e., responding to the visual target, when they require a common feature (Hommel et al., 2001; see also, Boulenger et al., 2006; Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a; Sato et al., 2008). Thus, we expected faster responses when the name and the following visual target did not share a common spatial feature (e.g., name of a superior protagonist, followed by a target below), relative to when the two events did share a common feature. Participants performed a visual discrimination task in Experiment 3, which we expected would be relatively more prone to interference from the name recognition task. In Experiment 4, participants

4 4 D. G. GOZLI ET AL. performed visual detection task, which we expected would be less prone to interference from the name recognition task (Gozli et al., 2013a). If slower responses on compatible trials were, indeed, caused by the namerecognition task interfering with visual discrimination, then the findings should be weak or absent in a detection task. This was confirmed in Experiment 4. Experiment 1 In this experiment, we tested the effectiveness of a set of descriptions (of other people) in producing spatial bias in an explicit localization task. Participants read about fictional undergraduate students who were either doing well (superior) or poorly (inferior) with regard to their academic performance, health or social life. Each participant read a total of three vignettes, each containing two protagonists. After reading a vignette, we used a pen-and-paper localization task, which required participants to assign the name of each protagonists to a location within a two-dimensional plane. Within this frame of reference, we measured deviations from the centre along both the horizontal and vertical axes. We predicted upward deviations for protagonists in a superior state and downward deviations for protagonists in an inferior state. Method Participants Forty-eight undergraduate students (29 female) at the University of Toronto gave informed consent and took part in this experiment in exchange for course credit. Prior to their participation, the participants were all unaware of the purpose of the study and this was confirmed in the debriefing phase (this was true for all four experiments). Moreover, given that we did not aim to address individual differences, in all the following experiments we did not collect any demographic information regarding the participants academic performance, health or social life. All experimental protocols were approved by the Research Ethics Board of the University of Toronto (protocol reference number: 26353; application title: How do concepts and perceptual simulations affect visual attention? ). Stimuli and procedure The experiment was conducted in dimly lit rooms. Male participants were given vignettes about male protagonists, and female participants were given vignettes about female protagonists. Each vignette described one superior protagonist and one inferior protagonist. Sample vignettes are included in Appendix 1 and the names used in the experiments appear in Appendix 2. The names used for superior and inferior protagonists were counterbalanced across participants. (e.g., for half of the male participants, Paul was the superior protagonist name while it was the inferior protagonist name for the second half of the male participants). Each vignette was divided into four pages. Participants were instructed to read the vignette carefully and remember as much information about the two protagonists as possible. They could navigate through the four pages using the left/right arrow keys. After reading each vignette, participants were presented with a single 8.5 by 11 (letter size) sheet of paper. On the paper, the two major axes (horizontal and vertical) were drawn, intersecting at the centre of the paper. The paper was otherwise empty. Participants were told that the study aimed to find out how we assign locations to people. They were asked to imagine themselves to be represented by the central location in the paper and assign a location to each of the two protagonists by writing the initial of each name (e.g., P for Paul ) somewhere on the sheet. Participants were given a chance to take a short break before reading the next vignette. Design Each participant read three vignettes on the topics of academic performance, health and social life (Appendix 1). Immediately after each vignette, participants performed the location assignment task, resulting in collecting six responses from each participant. The order in which the vignettes were presented to the participants was counterbalanced. Results and discussion Responses in the localization task were measured in millimetres in terms of their deviation from the horizontal axis (vertical deviation) and from the vertical axis (horizontal deviation). The averaged responses are presented in Figure 1, which clearly demonstrate a difference between responses to the inferior and superior protagonists. We submitted vertical deviations and horizontal deviations to two separate repeated-measures ANOVAs, with independent

5 VISUAL COGNITION 5 Figure 1. Result of the localization task in Experiment 1. Each participant read three vignettes about different aspects of student life (academic performance, health, social affiliations). Each vignette included two protagonists, one doing poorly (inferior protagonist) and one doing well (superior protagonist). The results show average deviation from the centre, along the horizontal and vertical axes, in millimetres. Error bars represent 95% within-subject confidence intervals. factors being vignette category (academic, health, social) and protagonist type (inferior vs. superior). For vertical deviations, we found no main effect of vignette category (F[2, 94] =.914), but a main effect of protagonist type (F[1, 47] = , p <. 001, h 2 p =.487) and an interaction between the two factors (F[2, 94] = , p <.001, h 2 p =.231). The interaction indicates different inferior-to-superior distance for the different vignette categories. Namely, the distance was largest for the academic vignette (85 mm, d z = 1.37), followed by the vignettes about social life (45 mm, d z =.66) and health (18 mm, d z =.22). For horizontal deviations, we again did not find a main effect of vignette category (F[2, 94] = 2.981, p =.056, h 2 p =.060), although we did find a main effect of protagonist type (F[1, 47] = 6.596, p =. 013, h 2 p =.123), with a statistically reliable distance along the horizontal axis between the superior and inferior protagonists (29 mm, d z =.37). There was no interaction between the two factors along the horizontal axis (F [2, 94] = 1.409, p =.249, h 2 p =.029). In both dimensions, the deviations among the responses for the two protagonists were in the expected direction. On average, the superior protagonists were assigned upward and right-ward locations, whereas the inferior protagonists were assigned downward and left-ward locations. The bias along the vertical axis was relatively more robust, which is consistent with previous findings that also indicate a more robust metaphorical mapping of valence/status onto the vertical domain, compared to the horizontal domain (e.g., de la Vega, De Filippis, Lachmair, Dudschig, & Kaup, 2012; Gozli et al., 2013b; Taylor et al., 2015). In addition to the possibly stronger metaphorical association, the bias along the horizontal axis has been shown to be sensitive to the participants handedness (e.g., Casasanto, 2009; de la Vega, Dudschig, De Filippis, Lachmair, & Kaup, 2013). Because of these factors, we do not investigate bias along horizontal axis in the following experiments. Rather, we investigate bias along the vertical dimension and note that the biases can, indeed, appear in an explicit localization task. The differences between the three vignettes can serve as a point of reference when evaluating performance in the following implicit tests of spatial bias. Experiment 2 Having shown the vignettes can produce spatial bias along the vertical dimension in an explicit task, our

6 6 D. G. GOZLI ET AL. goal for the second experiment was to demonstrate, using a variant of the spatial Stroop task, that thinking about the protagonists could induce a spatial bias. If recognizing the names of protagonists from the preceding vignette, without categorizing the names in terms of superior/inferior, can induce a spatial bias, this would be evidence that the participants applied spatial metaphors in thinking about the protagonists. After reading a vignette, in a separate task, the names of superior and inferior protagonists were presented above or below the fixation. The to-be-recognized name was presented at a location that was either metaphorically compatible or incompatible with what participants had read earlier about the person. In this paradigm, attentional orienting and name recognition are two attributes of the same event, because (a) the presentation of the name entails (i.e., partly consists of) the onset of a perceptual object above/below fixation, and (b) recognizing the name entails (i.e., partly consists of) having oriented one s attention to the name s location. Therefore, we expect compatibility to facilitate performance in this task. Method Participants Eighteen (six male) new undergraduate students at the University of Toronto gave informed consent and took part in this experiment in exchange for course credit. They all reported normal or correctedto-normal vision. Apparatus and stimuli The experiment was conducted in dimly lit rooms. Displays were presented on 19-inch CRT monitors set at resolution and 85 Hz refresh rate. The experimental programme was written in Matlab (MathWorks, Natick, MA), using the Psychophysics toolbox (Brainard, 1997; Pelli, 1997; version 3.0.8). Using a head/chin-rest, the distance from display was fixed at about 45 cm. The vignettes used were the same as in Experiment 1. In the present experiment, test names (Appendix 2) are those that appeared in the story (receiving yes responses in the recognition task), whereas catch names did not appear in the story (receiving no response in the recognition task). As before, the test names used for superior and inferior protagonists were counterbalanced across participants. Procedure Each vignette was divided into four screens. Participants were instructed to read the vignette carefully and remember as much information about the two protagonists as possible. They could navigate through the four screens using the left/right arrow keys. After finishing each vignette, participants performed 80 trials of a visual name recognition task. Each trial of the name recognition task began with a central fixation cross ( + ), presented at the display centre. After 1000 ms, a name appeared above or below the fixation cross (deviating 4 of visual angle). The question on each trial was whether the name belonged to the vignette they had just read. The name remained on display until a response was recorded. Participants pressed the / key for yes responses (correct response to test names) and the z key for no responses (correct response to catch names). Participants could receive three kinds of error feedback: if they responded within the first 100 ms following the onset of the name, they received anticipation-error feedback ( TOO QUICK! ); if they responded later than 1500 ms following the onset of the name, they received a delay error feedback ( TOO LATE! ); if they pressed the wrong key, they received an error feedback ( MISTAKE! ). All types of feedback would remain on display for 2000 ms. Design Each participant read three vignettes on the topics of academic success, health and social life (Appendix 1). Immediately after each vignette, participants performed the recognition task, which included the two (test) names from the vignette and two new (catch) names, resulting in a total of 120 test trials per participant (i.e., 40 test trials per vignette). An equal number of test names and catch names were used, and appearance of the two name types was equally probable, which means yes and no responses were equally likely. Test trials were further equally divided into those including superior and inferior names (respectively, the names of the superior and inferior protagonist). Both name locations (above vs. below) during the recognition task were equally likely, and name location varied orthogonally to protagonist category (superior vs. inferior) and recognition response (yes/no). Participants were given a chance to take a short break before reading the next vignette.

7 VISUAL COGNITION 7 Results and discussion Recognition accuracy was high on test trials (M ± SE = 93% ± 1%) and catch trials (91% ± 1%), with accuracy ranges of 83 97% and 78 98%, respectively, for test and catch trials. Catch trials were not analysed further. Before analysing the response times (RT) on test trials, incorrect responses and responses that fell 2.5 SD above or below the total mean were excluded (2.8% of trials). From the remaining data, mean RTs were submitted to a repeated-measures ANOVA, with vignette category (academic, health, social), protagonist type (superior vs. inferior) and name location (above vs. below) as factors (α =.05). The RT data are graphed in Figure 2. None of the main effects reached significance (vignette category, F[2,34] = 2.57, p =.091, h 2 p =.131, protagonist type, F [1.17] =.68, p =.42, h 2 p =.038, and name location, F [1.17] = 2.35, p =.144, h 2 p =.122). The two-way interactions between vignette category and protagonist type (F[2,34] =.15, p =.86, h 2 p =.009) and between vignette category and name location (F[2,34] = 2.01, p =.15, h 2 p =.106) also did not reach significance. We found, however, a two-way interaction between protagonist type and name location (F[1,17] = 6.43, p =.021, h 2 p =.274), which was not qualified by a three-way interaction (F[2,34] =.131, p =.88, h 2 p =.008). Consistent with the predicted metaphorical associations, responses were faster when names were presented at a metaphorically compatible location (523 ± 13 ms) compared to when they were presented at a metaphorically incompatible location (533 ± 14 ms, d z =.604). It should be noted, however, that this interaction was driven by test trials in which the superior protagonist was presented (518 ± 13 ms vs. 538 ± 16 ms, for compatible and incompatible trials, d =.62, p =.017), rather than trials with the inferior protagonist (531 ± 16 ms vs. 532 ± 14 ms, d =.046, p =.85). This finding is relevant when considering alternative explanations of the present findings in the General discussion. The interaction between protagonist type and name location, although relatively more robust at the level of the aggregate data (d z =.604), appears less robust within each vignette condition (see Figure 2). Two-way interaction effects at the level of each vignette were relatively weak (d z =.306,.371,.143, respectively, for the academic, heath, social-life vignette). It would be hasty to make distinctions between a vignette with which the null hypothesis was rejected vs. one with which the null was not rejected. When dealing with weaker effect sizes (i.e., here, the level of individual vignette), failing to reject the null on a subset of conditions is to be expected (e.g., Francis, 2012). Furthermore, the relatively weaker effects at the level of individual vignette are inconsistent with the idea that the results at the aggregate level were driven by data from one of the three vignettes. The relatively more robust interaction at the aggregate level might be due to individual differences in sensitivity to vignette types or some other source of noise at the level of individual vignettes. It is, therefore, important to keep in mind that our claims are made about the data at the aggregate level (Lamiell, 2000). A similar ANOVA run on percent error (PE) data did not reveal a significant interaction between protagonist type and name location (F[1, 17] =.374), ruling out a possible speed accuracy trade-off. A non-significant trend (F[1,17] = 2.82, p =.079, h 2 p =.171) indicated slightly lower PE when names were presented above (6% ± 1%) relative to when presented below fixation (8% ± 1%). None of the other main effects or interactions reached statistical significance (F values < 1). Consistent with our prediction, name-recognition performance varied systematically as a function of the relationship between protagonist type and name location. The pattern is consistent with the metaphorical compatibility effect, with the qualification that the compatibility pattern in this experiment was driven by the superior protagonists, which on average caused faster responses above fixation than below fixation. Unlike Experiment 1, the implicit test employed in the present experiment did not distinguish between the vignettes the three-way interaction was far from statistical significance. To provide converging evidence for the activation of the visuospatial features in thinking about the vignette protagonists, in Experiment 3 we used the attentional cueing task, which decouples name recognition from spatial orienting into two separate sub-tasks. Experiment 3 In this experiment, rather than presenting the name of the superior or inferior person above/below fixation, the name was presented at fixation. After presenting

8 8 D. G. GOZLI ET AL. Figure 2. Response time data from Experiment 2 (name recognition task) graphed as a function of vignette category (academic, health, social), protagonist (superior vs. inferior) and name location (above vs. below). Error bars represent 95% within-subject confidence intervals. the name, a target letter (X/O) was presented above or below fixation, such that the location of the visual target could be metaphorically compatible or incompatible with the metaphorical-spatial association with the protagonist. Thus, in this paradigm, name recognition and attentional orienting were aspects of two separate events. This has important consequences for performance, as when two consecutive cognitive events require the same feature, processing time is delayed relative to when the two events do not share any common feature. We employed a cueing method in which the cue (event 1) and the target (event 2) can be spatially compatible, while they otherwise remain perceptually incompatible. Estes, Verges, and Adelman (2015) and Ostarek and Vigliocco (2017) have demonstrated that cue target congruence can facilitate performance when there is both spatial and perceptual congruency between the two stimuli (e.g., BIRD followed by an image of a bird above fixation; SKY followed by an image of a cloud above fixation). Thus, if the two consecutive stimuli are sufficiently congruent across multiple dimensions, then the two stimuli can be represented as a single event and cue target compatibility yields facilitation. By contrast, if the two stimuli are perceptually dissimilar (e.g., BIRD followed by the letter X or O above fixation), then spatial compatibility yields interference (Amer et al., 2017; Estes, Verges, & Adelman, 2015; Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a; Ostarek et al., in press; Ostarek & Vigliocco, 2017). That is because the cue target dissimilarity prevents the two stimuli from being processed as a single event, and the partial compatibility of the two stimuli (i.e., a common feature) can interfere with performance. Such interference is consistent with the framework of the Theory of Event Coding (TEC; Hommel et al., 2001). According to TEC, once the common feature is bound to the representation of the first event, it is temporarily unavailable for the processing of the second event. In the present experiment, the two events are (a) activation of a protagonist s mental representation and (b) spatial orienting for the identification of the target letter. The possible common features in the two events are the spatial codes, up and down. If both events require the same spatial feature, responses should be delayed relative to when the two events require different spatial features (Estes, Verges, & Adelman, 2015; Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a; Gozli & Pratt, 2011; Ostarek & Vigliocco, 2017; Richardson, Spivey, Barsalou, & McRae, 2003). Therefore, metaphorical compatibility should lead to interference (slower RT on compatible trials compared to incompatible trials). Previous studies have demonstrated that the temporal delay (stimulus onset asynchrony; SOA) between the two event presentations can modulate the interaction between conceptual and spatial

9 VISUAL COGNITION 9 processing (e.g., Gozli et al., 2013a; Ostarek & Vigliocco, 2017). With implicitly spatial words, not associated with any context or referent (e.g., HAPPY, BIRD ), interference with visual discrimination can be observed with relatively shorter SOAs, about 300 ms (e.g., Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a; see also, Bergen, Lindsay, Matlock, & Narayanan, 2007). In the present study we do not use concepts that have pre-existing spatial associations, and each name is associated with a context and referent that is learned during the study session. Thus, based on the relatively richer and more concrete associations with the names, we considered the possibility that an interference can be observed with relatively longer SOAs ( ms). If both the spatial codes (up/down) remain occupied by the simulation of their corresponding (superior/inferior) protagonist, then responses in the metaphorically compatible trials (superior + above) should be delayed relative to when the incompatible trials. Method Participants Eighteen (seven male) undergraduate students at the University of Toronto gave informed consent and took part in this experiment in exchange for course credit. They all reported normal or corrected-to-normal vision, and they were all unaware of the purpose of the study. Apparatus, stimuli and procedure The apparatus and stimuli were identical to those in Experiment 2. In this experiment, however, after reading each vignette, participants performed 80 trials of a visual discrimination task. Each trial of this task began with the presentation of a fixation cross ( + ), which remained on display for 1000 ms. Next, one name replaced the fixation cross at the centre. After a delay (randomly selected from the interval between ms), a visual target (letter X or O ) was presented above or below the name (deviating by 4 of visual angle from the centre). Participants were instructed to respond to this visual target only if the central name belonged to the recently read vignette (using the / key for X and the z key for O ). If the name did not belong to the vignette, they were instructed to withhold responding (catch trial). Similar to Experiment 2, participants could receive three kinds of error feedback. If they responded within the first 100 ms following the onset of the name, they received anticipation-error feedback ( TOO QUICK! ). If they responded later than 1500 ms following the onset of the name, they received missed-trial feedback ( TOO LATE! ). In addition, if a participant responded on a catch trial, or if they responded with the wrong key on a test trial, they received a keypress-error feedback ( MISTAKE! ). All error feedback remained on display for 2000 ms. Design Similar to Experiment 2, each participant read three vignettes. Immediately after each vignette participants performed a name-recognition task, which included the two (test) names from the vignette and two new (catch) names. The appearance of the two name types were equally probable, resulting in a total of 120 test trials per participant (i.e., 40 test trials per vignette). Test trials were further equally divided into those including superior and inferior names. Target locations (above vs. below) were equiprobable and varied orthogonally to protagonist type (superior vs. inferior) and target letter ( X vs. O ). Participants were given a chance to take a short break before reading the next vignette. Results and discussion Accuracy was high on test trials (M ± SE = 94% ± 1%) and catch trials (99% ±.3%), with accuracy ranges of 77 98% and %, respectively, for test and catch trials. Catch trials were not analysed further. Before analysing the response times (RT) on test trials, incorrect responses and responses that fell 2.5 SD above or below the total mean (3.1% of trials) were excluded. From the remaining data, Mean RTs were submitted to a repeated-measures ANOVA, with vignette category (academic, health, social), protagonist type (superior vs. inferior) and visual target location (above vs. below) as factors (α =.05). The findings are graphed in Figure 3. None of the main effects reached significance (for vignette category, protagonist type, and target location, respectively, F[2,34] =.539, p =.59, h 2 p =.031, F[1.17] = 2.381, p =.141, h2 p =.123, and F[1.17] =.307, p =.59, h 2 p =.018). The twoway interactions between vignette category and protagonist type (F[2,34] = 1.07, p =.35, h 2 p =.059) and

10 10 D. G. GOZLI ET AL. Figure 3. Response time (RT) data from Experiment 3 (visual discrimination task) graphed as a function of name type (superior vs. inferior) and visual target location (above vs. below). Error bars represent 95% within-subject confidence intervals. between vignette category and target location also did not reach significance (F[2,34] = 1.74, p =.19, h 2 p =.093). Most importantly, we found an interaction between protagonist type and target location (F [1,17] = 24.42, p <.001, η 2 =.590). Responses were slower when targets were presented at a metaphorically compatible location (448 ± 12 ms) compared to when they were presented at a metaphorically incompatible location (429 ± 11 ms, d z = 1.22). The two-way interaction was present for both the superior protagonists (447 ± 11 ms vs. 425 ± 11 ms, respectively, for compatible and incompatible trials, d = 1.00, p <.001), and the inferior protagonists (450 ± 13 ms vs. 433 ± 12 ms, d =.62, p =.018), the implications of which will be discussed in considering alternative accounts in the General discussion. Finally, consistent with Experiment 2, the three-way interaction was far from statistical significance (F[2,34] =.274, p =.762, h 2 p =.016). Mean PEs were submitted to the same ANOVA, which did not reveal a significant two-way interaction between protagonist type and target location (F[1,17] = 2.64, p =.123, h 2 p =.134). Matching the RT findings, mean PE was higher on compatible trials (7% ± 1%) than the incompatible trials (5% ± 1%). None of the other main effects or interactions reached statistical significance (F < 1.93, p >.16). With regard to the temporal delay between the onset of the names and the onset of the visual targets, we should point out that previous research suggests that obtaining interference with overlearned stimuli (e.g., words referring to positive/negative valence or high/low power) requires short SOAs of around 300 ms (Estes, Verges, & Barsalou, 2008; Gozli et al., 2013a). The necessity of short SOAs with known words suggests that there is a brief period in which the spatial features are required for processing the words (i.e., the period in which the same spatial feature cannot participate in representing another event). By contrast, we found interference with relatively longer SOAs (average of 800 ms), which suggests a relatively longer temporal interval in which the spatial feature is engaged by thinking about the protagonists. Was the compatibility effect influenced by the cue target SOA? To address this question, we categorized SOAs into short ( ms) and long ( ms), then submitted the RT data to a repeatedmeasures ANOVA with cue target spatial compatibility (compatible vs. incompatible) and cue target SOA (short vs. long) as factors. This analysis found a main effect of compatibility (F[1,17] = 24.21, p <.001, h 2 p =.588), a main effect of SOA, (F[1,17] = 30.32, p <.001, h 2 p =.641), and a marginally significant interaction (F[1,17] = 4.19, p =.057, h 2 p =.198). The main effect of SOA reflects slower responses with short cue target delay (449 ± 11 ms) compared to long cue target delay (429 ± 11 ms). The interaction

11 VISUAL COGNITION 11 reflects a relatively smaller compatibility effect with short cue target delay (15 ms, d z =.81) compared to long cue target delay (24 ms, d z = 1.43). Thus, unlike several previous studies that found a diminishing interference effect with longer cue processing time (e.g., Gozli et al., 2013a; Lachmair et al., 2016; Ostarek & Vigliocco, 2017; see Ostarek et al., in press), we found an effect that was within the range of our experimental design larger with increased cue processing time. One may attribute this observation to the relatively rich and longlasting nature of mental simulations involved in thinking about the protagonists. Unlike tasks in which participants respond to single words, without a context in which those words acquire meaning, the cues in the present experiments are associated with relatively concrete contexts. The duration of time in which simulation of the context, in which the protagonists status is meaningful, might very well outlast the duration of time necessary for recognizing the name as belonging to the preceding vignette. As such, the present findings point out that methods that rely on the presentation of overlearned, out-of-context words might underestimate the effect and duration of mental simulation that occur when our thinking is embedded within more realistic situations. Rather than the timecourse of the interference effect, what was primarily important for us was the very fact that a visual interference can happen using the names referring to the superior/inferior protagonists. In Experiment 4, we attempted to test whether the interaction found in Experiment 3 was driven by interference on the metaphorically compatible trials. Experiment 4 The findings of Experiment 3, i.e., faster responses on metaphorically incompatible trials, are consistent with two interpretations. According to the first interpretation, thinking about the protagonist activated the associated spatial metaphor (e.g., superior is up), which then interfered with discriminating a visual target at a compatible location. When the two subtasks (or two distinct events) require a common feature (e.g., spatial feature up), the second sub-task slows down because the shared feature is temporarily integrated in, and thus occupied by, the first sub-task (Hommel et al., 2001; for similar accounts see, Estes, Verges, & Adelman, 2015; Ostarek & Vigliocco, 2017). According to a second interpretation, thinking about the protagonist orients the participants toward the metaphorically incompatible location (e.g., orienting downward after thinking about a superior protagonist). This would also slow down responses at the metaphorically compatible location, as a result of a bias in favour of the incompatible location and not as a result of a simulation-driven interference. We favour the first interpretation, although both are logically possible. Experiment 4 is an attempt to distinguish the two interpretations. If slower responses on compatible trials resulted from a spatial orienting toward the incompatible location, then this orienting would have generated a similar pattern of results for both a visual discrimination task and a visual detection task (e.g., Posner, 1980). On the other hand, if the slower responses on compatible trials resulted from interference at the compatible locations, i.e., difficulty in distinguishing the two possible targets, then the interference should be reduced or eliminated in a less demanding detection task, in which participants simply report the presence of a visual target (Gozli et al., 2013a, Experiment 3A). It is, indeed, possible that a simulationdriven interference would turn to a facilitation when we replace the discrimination task with a detection task or with identification of a visual target that fits into the simulation and thus resulting in the representation of a single event (Estes, Verges, & Adelman, 2015; Ostarek & Vigliocco, 2017; Ostarek et al., in press). We used a similar procedure as Experiment 3 but replaced the visual discrimination task with a visual detection task. Participants now pressed one key (spacebar) when they detected the presence of a visual target (always the letter O ) above or below the centrally presented name. The kind of bias that could facilitate discrimination responses (at the incompatible location) is a general spatial bias that would continue to facilitate responses in a detection task. Therefore, if the findings of Experiment 3 resulted from spatial orienting toward the incompatible locations, the present modification is likely to replicate the effect. However, if the findings of Experiment 3 resulted from a simulation-driven interference at the compatible locations, such a bias would likely cease to interfere with performance if the target feature is irrelevant, i.e., in detection task. Therefore, if the compatibility effect in Experiment 3 resulted from

12 12 D. G. GOZLI ET AL. interference at the compatible locations, then removing the necessity to discriminate the targets should reduce or eliminate the effect. Method Eighteen new undergraduate students participated in this control experiment. The experimental setup, procedure and design were identical to Experiment 3. We merely changed the visual discrimination task to a detection task, meaning that participants now pressed one key (spacebar) when they detected the presence of a visual target (always the letter O ) above or below the centrally presented name. Results and discussion Accuracy was high on test trials (M ± SE = 97% ± 1%) and catch trials (99% ±.2%), with accuracy ranges of 83 99% and %, respectively, for test and catch trials. Catch trials were not analysed further. Before analysing the RTs on test trials, incorrect responses and responses that fell 2.5 SD above or below the total mean (5.3% of trials) were excluded. Mean RTs were submitted to the repeated-measures ANOVA with vignette category (academic, health, social), protagonist type (superior vs. inferior) and visual target location (above vs. below) as factors (α =.05, see Figure 4). The analysis revealed a main effect of protagonist type (F[1,17] = , p =.002, h 2 p =.304), with faster responses after superior protagonists (M ± SE = 320 ± 11 ms), compared to inferior protagonists (M ± SE = 328 ± 11 ms). We found no main effect of target location (F[1,17] =.429), no interaction between protagonist type and target location (F [1,17] =.022), and no three-way interaction (F[1,17] =.275). Most important for our purpose was the absence of interaction between protagonist type and target location, performance was similar on compatible (322 ± 12 ms) and incompatible trials (323 ± 10 ms). Given the null results in the present experiment, and given our aim to use these results in achieving a better understanding of the results of Experiment 3, we ought to include the two sets of data in one analysis (e.g., Nieuwenhuis, Forstmann, & Wagenmakers, 2011). The null findings of Experiment 4 would be more informative in light of a three-way statistical interaction between task (discrimination vs. detection), protagonist type (superior vs. inferior), and target location (above vs. below). Thus, we submitted the data to a mixed ANOVA, with vignette category, protagonist type, target location as within-subject factors, and task as the betweensubject factors. This analysis revealed a two-way interaction between protagonist type and target location, F (1,34) = 9.042, p =.002, h 2 p =.210, which was qualified by a three-way interaction between protagonist type, target location, and task, F(1, 34) = 10.38, p =.003, h 2 p =.234. Therefore, the analysis confirmed the difference in the compatibility effects across Experiments 3 and 4. The findings do not favour a facilitation account (i.e., attentional orienting toward the incompatible location), because such a facilitation would persist with the visual detection task (Gozli et al., 2013a, Experiments 1A and 1B). Instead, the null effect in Experiment 4 suggests that the findings of Experiment 3 were more likely driven by a simulation-driven interference at the metaphorically compatible location. We would expect such an interference to be more robust in the case of the relatively more demanding discrimination task. What about the possible role of cue target SOA in modulating spatial compatibility? Same as Experiment 3, we submitted the RT data to a separate repeatedmeasures ANOVA, with cue target compatibility and SOA ( short [ ms] vs. long [ ms]) as factors. This analysis revealed no main effect of compatibility (F[1,17] < 1), although it did reveal a main effect of SOA (F[1,17] = 37.64, p <.001, h 2 p =.689). The two-way interaction did not reach significance (F[1,17] = 2.69, p =.120, h 2 p =.136). The main effect of SOA reflects slower responses with short cue target delay (337 ± 12 ms) compared to long cue target delay (312 ± 10 ms). Numerically, the RT difference (incompatible compatible) would be consistent with interference at short SOA ( 6 ms,d z =.19) and with facilitation at long SOA (9 ms, d z =.38). Therefore, the results of Experiment 4 do not in any way indicate a bias in favour of the incompatible locations. The dissimilarity of the results of Experiments 3 and 4 speak against another alternative account of Experiment 3, which is based on inhibition of return (IOR). An IOR account would propose participants initially orient to the compatible location and then inhibit it, leading to an advantage for the incompatible location